1 JANUARY 2005
BIASUTTI ET AL.
211
Terrestrial Influence on the Annual Cycle of the Atlantic ITCZ in an AGCM Coupled
to a Slab Ocean Model*
M. BIASUTTI, D. S. BATTISTI,
AND
E. S. SARACHIK
Department of Atmospheric Sciences, University of Washington, Seattle, Washington
(Manuscript received 29 September 2003, in final form 14 July 2004)
ABSTRACT
An atmospheric GCM coupled to a slab ocean model is used to investigate how temperature and
precipitation over South America and Africa affect the annual cycle of the Atlantic ITCZ. The main
conclusion of this study is that variations in precipitation and temperature forced by the annual cycle of
insolation over the continents are as important as variations in insolation over the ocean and in ocean heat
transport convergence in forcing the annual march of the Atlantic ITCZ observed in the control simulation.
The processes involved are as follows.
The intensity of precipitation over land affects the stability of the atmosphere over the tropical Atlantic
Ocean, and thus modulates the intensity of deep convection and convergence in the ITCZ. Both the
imposed changes in land precipitation and the subsequent changes in the strength of the ITCZ drive surface
wind anomalies, thereby changing the meridional gradient of SST in proximity of the basic-state ITCZ.
Finally, atmosphere–ocean feedbacks cause the ITCZ to be displaced meridionally.
Seasonal changes in surface temperature in the Sahara also have a strong influence on the position of the
Atlantic ITCZ. Cold wintertime temperatures produce high surface pressure anomalies over Africa and into
the tropical North Atlantic and drive stronger trade winds, which cool the North Atlantic by evaporation.
The coupled interactions between the SST, the wind, and the ITCZ intensify the anomalies in the equatorial
region, causing the southward displacement of the ITCZ in boreal spring.
1. Introduction
This is the third in a series of papers aimed at decomposing the annual cycle of precipitation in the
tropical Atlantic sector into its constituent elements. In
Biasutti et al. (2003, hereafter BBS1) and Biasutti et al.
(2004, hereafter BBS2) we have described a set of atmospheric general circulation model (AGCM) experiments that separate the annual cycle at any given location into “locally” and “remotely” forced components.
These papers described how the annual cycle over land
influences the annual cycle of the Atlantic ITCZ when
SST is fixed, and how the annual cycle of SST influences
precipitation over the continents, in equatorial Africa
and equatorial South America, and in the Sahel.
Our strategy for identifying the influence of SST on
the continental annual cycle of temperature and precipitation and the influence of the continents on the
* Joint Institute for the Study of the Atmosphere and Oceans
(JISAO) Contribution Number 1005.
Corresponding author address: Michela Biasutti, LamontDoherty Earth Observatory of Columbia University—Oceanography, 61 Route 9W, P.O. Box 1000, Palisades, NY 10968-8000.
E-mail: [email protected]
© 2005 American Meteorological Society
JCLI3262
maritime annual cycle of precipitation was to treat the
SST and the insolation boundary conditions as independent forcings. So, for example, the influence of SST on
the continental climate was inferred from a simulation in
which insolation was kept fixed at March value but SST
cycled through the observed climatology: in such an experiment, any annual variations of the continental climate
must be solely due to the remote influence of the SST.
Similarly, the reverse experiment—in which SST was kept
fixed at March value but insolation cycled through the
observed climatology—illuminated the influence of continental climate on the oceanic climate in an AGCM.
One limitation of our previous work is obvious: the
terrestrial influence on the oceanic climate is artificially
constrained by the specification of SST. In this study,
we couple a slab ocean model (SOM) to the AGCM so
that SST can respond to the terrestrial forcing via thermodynamic processes (i.e., in response to changes in
surface heat fluxes), albeit not via ocean dynamics. This
is still a strong limitation, especially in the equatorial
region, where Ekman transport and upwelling force the
annual appearance of the equatorial cold tongue. This
study will quantify the first-order effect of ocean dynamics in forcing the annual cycle of SST and of the
oceanic ITCZ, but will mostly focus on identifying the
mechanisms that are important for the development of
212
JOURNAL OF CLIMATE
the annual cycle in the Atlantic in the absence of interactive ocean dynamics. It should primarily be considered another step toward understanding how our fully
coupled models work.
Three considerations spur our interest in understanding the mechanisms by which the annual cycle is simulated in an AGCM⫹SOM model configuration. First,
current coupled general circulation models do a poor
job at simulating the climate of the Atlantic without
using flux adjustments (Davey et al. 2002). In order to
identify which processes might be responsible for this
failure, it is wise to look at a simpler model that can
reproduce the annual cycle of the Atlantic climate and
identify the mechanisms controlling the annual cycle in
this simplified context. That knowledge can then be a
starting point for diagnosing the source of error in more
complete coupled models. Second, we anticipate that
mechanisms important for the annual cycle in the tropical Atlantic are also relevant to interannual variability
(Hastenrath 1984). AGCM⫹SOM configurations have
been widely and successfully used to simulate the Atlantic interannual variability (e.g., Seager et al. 2001;
Saravanan and Chang 1999), indicating that any insight
on the annual cycle gained from the same configuration
might be applicable to the interannual variability problem. Third, understanding the first-order response of
the tropical Atlantic sector to large changes in external
forcing—such as changes in local insolation—can offer
guidance in organizing the climate information in the
paleo records from the Atlantic Ocean, Africa, and
South America into a cohesive picture.
This paper is organized as follows: Section 2 introduces the model and details the experimental design.
Section 3 describes the response of the ITCZ to a prescribed change in continental convective heating. Section 4 describes the remote influence of the annual
cycle in land surface temperature and land precipitation
on the seasonal cycle of the maritime ITCZ. Section 5
describes the local response of the ITCZ to the annual
cycle of insolation over the ocean and the annual cycle
of ocean heat transport convergence. Section 6 summarizes the paper, draws a comparison of the response of
the tropical Atlantic to local and remote forcings, and
discusses the implications of our results.
2. Experimental design
We use the Community Climate Model version 3
(CCM3) AGCM at its standard resolution (T42 with 19
VOLUME 18
vertical levels), coupled in the tropical Atlantic region
to a fixed 50-m-depth ocean mixed layer. This configuration allows for thermodynamic atmosphere–
ocean interactions, but no ocean dynamics. A detailed
description of the atmospheric model and a comparison of the simulated climate with observations can be
found in Kiehl et al. (1998), Hack et al. (1998), and
other articles in the same special issue of Journal of
Climate (1998, vol. 11, no. 6). Chang et al. (2000) and
BBS1 have looked in more detail at how CCM3 simulates the climate of the tropical Atlantic in the uncoupled configuration. Saravanan and Chang (2000)
and Chiang et al. (2003), each using a configuration
similar to the one used in this study, have shown that
CCM3 coupled to a slab ocean model simulates the
mean state and the variability of the tropical Atlantic
climate fairly well.
The effect of ocean dynamics on the annual cycle of
SST is parameterized in the SOM by specifying an annually varying heat flux correction (referred to as the Q
flux). The Q flux is derived from an uncoupled simulation that used fixed monthly mean observed SST; it is
calculated as the difference between the mixed layer
heat content tendency and the heat flux provided by the
atmosphere (the sum of turbulent surface fluxes and
radiative fluxes). The application of the Q flux can be
interpreted as the imposition of the ocean heat transport convergence into the oceanic mixed layer, although in reality it also corrects for biases in the atmospheric fluxes. Note that the Q flux is calculated from
monthly mean SST tendency and atmospheric fluxes,
and this mean value is then linearly interpolated to obtain the instantaneous values used in the integration.
This procedure introduces a small additional bias in the
simulated climatology.
We couple CCM3 to the SOM only in the tropical
Atlantic region and impose climatological SST elsewhere, smoothing the transition over 10° of latitude.
This configuration permits us to observe the response
of the tropical Atlantic to a forcing over Africa and
South America, while at the same time minimizes the
influence of the midlatitudes and of other oceans on the
tropical Atlantic climate.
Tables 1 and 2 give a succint overview of the uncoupled and coupled experiments presented in this paper. A more expansive summary is given in the appendix. Recall that in our effort to decompose the annual
cycle of the Atlantic ITCZ into locally forced and re-
TABLE 1. Uncoupled experiments: name, insolation forcing over land, SST boundary conditions, and elevated condensational
heating over South America and Africa.
Name
Insolation over land
SST
Land condensational heating
CTL (Control)
PM (Perpetual Mar)
PMS (Perpetual Mar SST)
PMJQ (Perpetual Mar with Jun heating)
LPVE (Perpetual vernal equinox over land)
Climatological
Vernal equinox (21 Mar)
Climatological
Vernal equinox (21 Mar)
Vernal equinox (21 Mar)
Climatological
Mar
Mar
Mar
Climatological
Calculated (⫽Climat)
Calculated (⬃Mar)
Calculated (⬃Climat)
Prescribed (⫽Jun)
Calculated (⬃Mar)
1 JANUARY 2005
213
BIASUTTI ET AL.
TABLE 2. Coupled experiments: name, insolation forcing over land, Q flux, and elevated condensational heating over South
America and Africa.
Name
Land insolation
Ocean insolation
Q flux
Land condensational heating
CpldCTL
CpldPJQ
CpldPMQ
CpldLPVE
CpldLPVEQflux
CpldLOPVE
Climatological
Climatological
Climatological
Vernal equinox (21 Mar)
Vernal equinox (21 Mar)
Vernal equinox (21 Mar)
Climatological
Climatological
Climatological
Climatological
Climatological
Vernal equinox (21 Mar)
Climatological
Climatological
Climatological
Climatological
Annual mean
Climatological
Calculated (⫽Climat)
Prescribed (⫽Jun)
Prescribed (⫽Mar)
Calculated (⬃Mar)
Calculated (⬃Climat)
Calculated (⬃Mar)
motely forced processes, we treat the ocean forcings
and the land forcings as independent. Accordingly, our
simulations are distinguished by independently considering the annual cycle of insolation over land, SST (for
the uncoupled simulations), and insolation over the
ocean and Q flux (for the coupled simulations).
In most simulations the elevated condensation heating is calculated by the convective parameterization in
CCM3. However, in selected simulations (see Tables 1
and 2 and the appendix), we have discarded the condensation heating over Africa and South America calculated by the model and instead prescribed a constant
value (obtained from a previous simulation corresponding to either June or March insolation values). By prescribing the continental condensation heating, we decouple continental convection from continental surface
temperature and can observe the response of the ocean
to each forcing separately.
In the following sections we will show differences
between a control simulation, in which insolation and
ocean Q flux vary climatologically, and simulations in
which the annual cycle has been totally or partially suppressed over either the landmasses or the oceans. For
example, the difference between the control simulation
and an experiment with perpetual boreal vernal equinox insolation over land (coupled control minus
coupled land perpetual vernal equinox, CpldCTL–
CpldLPVE) shows the ocean response to the annual
cycle over land induced by insolation. Similarly, the
difference between the control simulation and an experiment with climatological insolation and Q flux but
prescribed fixed elevated condensation heating over
Africa and South America (coupled control minus
coupled perpetual March heating, CpldCTL–CpldPMQ)
shows the ocean response to the annual cycle of precipitation over land, in the absence of an annual cycle of
land surface temperature.
In the rest of the paper we will briefly reintroduce
our experiments and explain how we interpret the differences among them. The reader is directed to Tables
1 and 2 and the appendix for a comprehensive reference to all the experiments and the differences in the
applied forcings that distinguish them.
All simulations that have an annual cycle in the
forcing were run for at least 12 years; the first 4 years
were discarded, and the analysis was conducted on
the average of the remaining years. Simulations in
which all forcings were kept constant were run for 12
months.
3. The ITCZ response to a steady
continental forcing
When SST is prescribed underneath an AGCM, the
response of the simulated oceanic ITCZ to annual
changes in insolation over the adjacent continents is a
change in intensity, but not in position (BBS1). Sensitivity experiments forced by continental elevated heating (and analyzed in detail in BBS2) suggest that, in
uncoupled simulations, continental precipitation modulates the intensity of maritime precipitation. The insolation-induced changes in continental convection drive
changes in the stability (temperature) of the free troposphere throughout the tropical Atlantic, and thus affect the intensity of convection; the changes in land
surface temperature do not appreciably affect the oceanic precipitation directly (BBS2). What is the response
of the ITCZ to continental forcings when we expand
our model to include thermodynamic atmosphere–
ocean interactions? What other mechanisms come into
play? We approach this question in steps: First, we investigate the response of the Atlantic SST and ITCZ to
a steady elevated condensation heating imposed over
South America and Africa (this section). Second, we
show the maritime response to insolation-induced annual variations in land surface temperature and precipitation over the continents (section 4).
Figure 1 shows the forcing applied in all experiments
described in this section. The imposed forcing is the
difference in condensation heating over South America
and Africa between the months of June and March [as
simulated by CCM3 in an experiment with prescribed
March SST and climatologically varying insolation, denoted in BBS1 by perpetual March SST (PMS)]. Figure
1a shows the horizontal pattern of the heating difference at 500 mb; Fig. 1b shows the vertical profile of the
heating difference averaged over South America.
Figure 2a shows the oceanic response in precipitation
and surface winds to the forcing shown in Fig. 1 in an
uncoupled simulation with March SST and insolation
conditions (we also performed a similar set of simulations, but with September conditions, which supports
the conclusions drawn in this section). Specifically, Fig.
214
JOURNAL OF CLIMATE
VOLUME 18
FIG. 1. The Jun–Mar difference in condensational heating over South America and Africa, taken
from a simulation with perpetual Mar SST and climatologically varying insolation over land. (a) Heating
difference at 500 mb. The contour interval is 2 ⫻ 10⫺5 K s⫺1, starting with ⫾10⫺5. Negative contours are
dashed, the zero line is omitted, shading indicates values larger than 10⫺5 K s⫺1. (b) Vertical profile
as a function of pressure (mb) of the heating difference averaged over South America (in units of
10⫺5 K s⫺1).
2a shows the “PMJQ minus PM” difference, where PM
simulates a perpetual March and PMJQ simulates a
perpetual March with imposed June condensation heating over the continents. The mechanisms responsible
for the uncoupled response have been described in
BBS2 and can be summarized as follows. A prescribed
negative heating over the continents causes a cooling of
the free troposphere that is spread horizontally by the
atmospheric circulation, thereby changing the stability
profile of the atmosphere everywhere in the Tropics
outside the area of prescribed heating. In the Atlantic
ITCZ region, CAPE is enhanced and precipitation becomes more vigorous. In regions that are nonconvective in the basic state, the change in stability is not
sufficient to induce deep convection, and thus the precipitation response is muted outside of the ITCZ region.
Figures 2a and 2b show that the surface wind response is more intense near the equator and to the
southern edge of the precipitation anomalies. The
analysis of sensitivity experiments (not shown) in which
the elevated heating is applied only in South Africa or
only in South America suggests that the wind anomalies
shown in Fig. 2a should be interpreted as mostly due to
both the Rossby wave response to African heating and
the feedback associated with changes in the ITCZ itself;
the wind response to South American heating is small
by comparison.
How does the coupling affect the ITCZ response?
Figure 2b shows the difference in precipitation and surface winds between the CpldPJQ and CpldPMQ simulations, that is, between coupled AGCM⫹SOM simulations in which condensation heating over Africa and
South America is prescribed to be fixed at the June and
March values, respectively, but otherwise the climate is
forced by the annual cycle of insolation and Q flux.
Therefore, the difference “CpldPJQ minus CpldPMQ”
is due to the difference in the continental condensation
heating shown in Fig. 1, but the coupled simulations
have an annual cycle in their basic state. Figure 2b
shows the CpldPJQ⫺CpldPMQ difference during
March, to be compared with Fig. 2a.
As in the uncoupled case, when continental precipitation is decreased, the intensity of precipitation in the
ITCZ is enhanced (by roughly 20%, when averaged
over the whole ITCZ). However, the coupled response
is drastically different from the uncoupled response in
that, in the coupled case, the ITCZ has shifted 7° northward in response to the imposed continental condensation heating anomalies. This suggests that precipitation
over the African and South American landmasses has
the ability to affect the annual meridional march of the
ITCZ.
The northward shift of the ITCZ could have been
anticipated in light of the surface wind response in the
uncoupled case (Fig. 2a): the southeasterly trades to the
south of the ITCZ are anomalously strengthened in
response to the continental forcing. Stronger winds
mean stronger evaporation and latent heat loss from
the ocean. When a slab ocean model is coupled to the
AGCM, this anomalous latent heat flux cools the SST
to the south of the mean ITCZ, establishing a crossequatorial gradient that “pushes” the ITCZ northward.
Associated with the SST gradient and the ITCZ shift is
an enhanced meridional wind response (cf. Figs. 2b
with 2a). The mechanisms by which the surface wind,
and thus the surface convergence and the ITCZ, are
thought to respond to a meridional SST gradient
anomaly have been described, for example, by Lindzen
and Nigam (1987) and Hastenrath and Greischar
1 JANUARY 2005
BIASUTTI ET AL.
FIG. 2. The surface wind and ITCZ response forced over the
ocean by the constant condensational heating anomalies shown in
Fig. 1. In experiments with prescribed SST, the response of the
Atlantic ITCZ is more intense precipitation, and the wind response is concentrated to the south of the ITCZ. In experiments
with interactive SST, the ITCZ is both more intense (by about
20%) and shifted to the north. Associated with the northward
shift of the ITCZ is a strong southerly wind anomaly. (a) Perpetual Mar with Jun continental elevated heating minus perpetual
Mar (PMw/JQ ⫺ PM). (b) Perpetual Jun continental elevated
heating expt minus perpetual Mar continental elevated heating
expt (CpldPJQ ⫺ CpldPMQ) for Mar insolation and SST conditions. The contour interval is 4 mm day⫺1, starting with the ⫾2
contour, negative contours are dashed, the zero line is omitted,
shading denotes anomalies larger than 2 mm day⫺1, and only wind
anomalies larger than 1 m s⫺1 are plotted. Anomalies over land
have been masked out.
(1993). Very briefly, SST anomalies are communicated
by turbulent mixing to the boundary layer, thus creating a low-level anomaly in pressure, wind, and convergence. This model works quite well in regions of strong
gradients, but surface winds and convergence are also
influenced by the free troposphere and thus feel the
effect of the elevated condensational heating released
in the ITCZ itself (McGauley et al. 2004; Chiang et al.
2001; BBS2). Therefore, it is best to think of the SST,
the surface wind, and the ITCZ as a coupled system.
Note that over the ocean the wind response to the
continental elevated heating forcing of Fig. 1 is comprised of two components: the direct response to con-
215
tinental heating (Gill 1980), which we expect to be constant, and the response to changes in oceanic precipitation, which we expect to depend on the position of the
basic-state ITCZ. It follows that a constant anomaly in
continental precipitation does not force a constant precipitation anomaly over the ocean but, rather, a constant meridional displacement of the ITCZ from its basic-state annual cycle.
This can be seen in Fig. 3, which shows the annual
cycle of the CpldPJQ⫺CpldPMQ anomalies in precipitation, SST, surface winds and heat flux into the surface
ocean in the central Atlantic (30°W) as a function of
month and latitude. Also shown is the annual cycle of
the position of the ITCZ (in this case identified by
the confluence line) in both runs. The ITCZ in the
CpldPJQ run is always about 7° to the north of that in
the CpldPMQ run with the largest displacement of the
ITCZ rainfall occurring in boreal spring, the smallest in
boreal winter (Fig. 3a). The ITCZ displacement is associated with a positive anomalous meridional SST gradient and wind anomalies that maximize to the north of
the CpldPMQ confluence line (Fig. 3b); outside of the
region spanned by the annual march of the ITCZ the
annual cycle of the anomalies is minimal. A band of
strong heat flux anomalies (Fig. 3c) follows the annual
march of the ITCZ. (Note that, although in this model
SST anomalies can only be generated by heat flux
anomalies, the two fields are not in perfect correspondence. The annual mean of the heat flux anomalies is
zero by construction, as we are analyzing the equilibrium state; the equilibrium annual mean SST anomalies
need not be—and are not—zero.)
Figure 3d shows the CpldPJQ⫺CpldPMQ anomalies
in wind speed squared; because evaporation depends
on the square of the wind, the similarity in the patterns
in Figs. 3c and 3d suggests that heat flux anomalies can
be explained by wind-induced anomalies in evaporation
(a larger wind speed means more evaporation and a
cooling of the ocean). A decomposition of the total
heat flux anomalies confirms that the wind-induced latent heat flux anomalies are the dominant component
of the strong heat flux anomalies in the equatorial region, although radiative flux anomalies (not shown) are
also nonnegligible. This finding supports our original
supposition: the marked annual cycle in the response of
SST to a steady continental forcing is due to the fact
that wind changes depend on the basic-state ITCZ.
To summarize, a reduction in continental precipitation induces a remote cooling of the free troposphere
and a reduction in stability. In the ITCZ, this translates
in more active deep convection and larger low-level
convergence. The surface wind anomalies (generated in
response to both the prescribed elevated heating over
land and the precipitation changes in the ITCZ) induce
a cooling to the south of the mean ITCZ (Figs. 2a,b).
Finally, SST, surface wind, and oceanic precipitation
adjust to each other in an interactive way, establishing
the anomalies portrayed in Figs. 2b and 3. A constant
216
JOURNAL OF CLIMATE
VOLUME 18
anomaly in continental precipitation forces a nearly
constant meridional displacement of the ITCZ from its
basic-state annual cycle.
4. The ITCZ response to an annually varying
continental forcing
a. Response to insolation-induced changes in land
surface temperature and precipitation
In this section we present the response of the Atlantic ITCZ to annually varying forcing from the continents. Specifically, we compare two simulations: a control run, with climatologically varying insolation and Q
flux, and an experiment in which insolation over land is
fixed at the boreal vernal equinox, while insolation over
the ocean and the Q flux vary climatologically. The
difference between the two simulations is the response
to the top-of-the-atmosphere insolation differences
shown in Fig. 4. It can be thought of as the response to
the annual cycle of insolation over land.
Before presenting the results of the coupled experiments, we will review the analogous uncoupled experiments (i.e., with prescribed SST). Figures 5a, 5c, 5e, and
5g show the difference between the uncoupled CTL
simulation and the uncoupled LPVE simulation (forced
with perpetual vernal equinox insolation over land)
during March, June, September, and December. Thus,
the precipitation anomalies shown in Fig. 5 (left) are
forced by the corresponding difference in insolation
over land from the equinox value (Fig. 4) and are somewhat modified by soil processes, which hold memory of
the forcing from one month to the next (BBS1). Figures
5a, 5c, 5e, and 5g present a companion view of the
insolation-induced anomalies that were presented in
BBS1 (cf. Fig. 10 of that paper). The left column of Fig.
5 shows that, over land, to zero order precipitation increases with increasing insolation and that, over the
ocean, the intensity of the ITCZ is sensitive to conti-
←
FIG. 3. The annual cycle of the response of the central Atlantic
(30°W) to the forcing in land condensational heating shown in Fig.
1, as a function of month and latitude [perpetual Jun continental
elevated heating minus perpetual Mar continental elevated heating (CpldPJQ ⫺ CpldPMQ)]. (a) CpldPJQ ⫺ CpldPMQ precipitation anomalies (contour interval of 4 mm day⫺1, starting with
⫾2). (b) CpldPJQ ⫺ CpldPMQ SST anomalies (contour interval
of 0.5°C; the zero line is omitted) and surface wind anomalies
(only wind anomalies larger than 2 m s⫺1 are plotted; the bold
arrow in the upper-left corner corresponds to a 10 m s⫺1 wind
velocity). (c) CpldPJQ ⫺ CpldPMQ total heat flux anomalies into
the ocean (contour interval of 30 W m⫺2; the zero line is omitted).
(d) CpldPJQ ⫺ CpldPMQ wind speed squared anomalies (contour interval of 16 m2 s⫺2; the zero line is omitted). In every panel
the thick lines indicate the climatological position of the confluence line in CpldPJQ (farther to the north) and CpldPMQ (farther to the south), the dashed lines indicate negative values, and
the shading indicates (b),(c),(d) positive values or (a) values
larger than the value of the first positive contour line.
1 JANUARY 2005
BIASUTTI ET AL.
217
nental forcing, but, when SST are prescribed, the location of the ITCZ hardly changes, at the resolution of
these simulations. We have suggested in BBS2, and reviewed in section 3, that the changes in ITCZ intensity
are a response to the changes in the free troposphere
stability that accompany changes in continental precipitation. The concomitant changes in land surface temperature do not have a sizable direct effect on oceanic
precipitation in the uncoupled experiment (BBS2).
Figures 5b, 5d, 5f, and 5h show the CpldCTL⫺
CpldLPVE difference in precipitation during March,
June, September, and December. The forcing is the
same as that for the uncoupled case (Fig. 4), but now
SST is allowed to change in response to changes in
turbulent and radiative surface heat fluxes. As expected, the precipitation anomalies over land are very
similar to those from the uncoupled case (cf. the left
and right columns of Fig. 5). Over the ocean, the precipitation anomalies are similar to those in the uncoupled case during June, September, and December
but show a markedly different pattern during March. In
the uncoupled case (Fig. 5a), the March anomalies
show a slightly weakened ITCZ. In the coupled case
(Fig. 5b), the ITCZ is both weakened and shifted farther south in the CpldCTL simulation compared to the
CpldLPVE simulation.
Figure 6 provides a better description of the influence that the seasonal cycle in insolation over land has
on the seasonal cycle over the tropical ocean. It shows
the annual cycle of precipitation and SST in the central
Atlantic (at 30°W) in the CpldCTL and the CpldLPVE
simulation. In the control simulation (Fig. 6a), the main
ITCZ stays north of 5°S and a secondary precipitation
center develops during boreal spring at about 10°S. In
the CpldLPVE experiment, in which the annual cycle
of both temperature and precipitation over the continents is suppressed (Fig. 6b), the northern ITCZ does
not reach as far south, while the southern precipitation
develops earlier in the year and is found farther south
at about 20°S. We see, by comparing Figs. 6a and 6b,
that suppressing the annual cycle over the continents
markedly suppresses the annual cycle in the position of
the ITCZ.
The mechanisms by which land convective heating
can affect the position of the ITCZ have been addressed in section 3. In this section we will focus our
description on how continental surface temperatures
can affect the ITCZ. We will analyze the CpldCTL–
CpldLPVE differences, which compound the effect of
←
FIG. 4. The annual cycle of TOA insolation over land, shown as
differences from the (boreal) vernal equinox value. (a) Mar
CTL ⫺ Mar LPVE. (b) Jun CTL ⫺ Jun LPVE. (c) Sep CTL ⫺
Sep LPVE. (d) Dec CTL ⫺ Dec LPVE. The contour interval is 4
W m⫺2 (40 W m⫺2), starting with ⫾2(20) in (a) and (c) [(b) and
(d)]. The dashed contours indicate negative values and the shading indicates positive values larger than 2 W m⫺2 (20 W m⫺2).
218
JOURNAL OF CLIMATE
VOLUME 18
1 JANUARY 2005
BIASUTTI ET AL.
219
FIG. 6. The annual cycle of precipitation, SST, and surface wind in the central Atlantic
(30°W) in simulations with different annual forcings. (a), (c) CpldCTL: annual cycle due to
both local (insolation and Q flux) forcings and remote (land insolation) forcing. (b), (d)
CpldLPVE: annual cycle due only to local forcings (insolation and Q flux); the remote forcing
coming from the annual cycle over land has been suppressed. The contour interval for precipitation is 3 mm day⫺1, starting with ⫾1.5; values larger than 4.5 mm day⫺1 are shaded. The
contour interval for SST is 1°C; values larger than 27°C are shaded. Only surface wind vectors
larger than 5 m s⫺1 are plotted; the bold arrow in the lower-left corners indicates a 10 m s⫺1
wind speed.
both land elevated heating and land surface temperature, but the effect of temperature is dominant (cf. section 4b) and can be easily isolated in the analysis.
Figure 7 shows the spatial pattern of CpldCTL–
CpldLPVE surface air temperature (SAT) difference
during March over the entire Atlantic sector (over the
ocean SAT closely mimics SST). To help locate the
precipitation displacement due to continental forcing,
we have plotted the zero line in precipitation anomalies
from Fig. 5b in Fig. 7. Again, it is clear that the southward displacement of the ITCZ is associated with an
anomalous negative meridional gradient of SST (and
hence surface air temperature) near the equator. How
is this anomalous gradient established?
Figure 8 shows the anomalies in total surface heat
flux (Fig. 8a) and latent surface heat flux (Fig. 8b) integrated over December, January, and February. The
latent heat flux makes up the bulk of the total heat flux
in the equatorial area, but radiative fluxes are also important, especially in the stratus region off the South
African coast. The similarities between Figs. 7 and 8a in
the deep Tropics and between Figs. 8a and 8b suggest
that wintertime December–February (DJF) latent heat
flux anomalies are responsible for the generation of the
anomalous equatorial SST gradient during March1—
and thus for the shift of the ITCZ. (When comparing
1
We remark that the anomalous equatorial meridional SST
gradient is negligible prior to the DJF season (not shown).
←
FIG. 5. The precipitation response to the annual cycle of TOA insolation over land in the case of (left column)
prescribed SST and (right column) interactive SST. The effect of having an interactive SST is to allow the ITCZ
to move, as can be easily inferred by comparing the coupled and uncoupled Mar anomalies. (a) Mar CTL ⫺ Mar
LPVE, (b) Mar CpldCTL ⫺ Mar CpldLPVE, (c) Jun CTL ⫺ Jun LPVE, (d) Jun CpldCTL ⫺ Jun CpldLPVE, (e)
Sep CTL ⫺ Sep LPVE, (f) Sep CpldCTL ⫺ Sep CpldLPVE, (g) Dec CTL ⫺ Dec LPVE, (h) Dec CpldCTL ⫺ Dec
CpldLPVE. The contour interval is 4 mm day⫺1, starting with ⫾2; the dashed contours indicate negative values;
and the shading indicates positive values larger than 2 mm day⫺1.
220
JOURNAL OF CLIMATE
FIG. 7. The surface air temperature response to the annual cycle
of insolation over land in the case of interactive SST (Mar
CpldCTL ⫺ Mar CpldLPVE). The contours are spaced logarithmically: ⫾0.5°, ⫾1°, ⫾2°, ⫾4°, ⫾8°C; negative values are dashed;
and positive values larger than 0.5°C are shaded.
Figs. 7 and 8, recall that poleward of 20° the calculated
SST is linearly merged with the prescribed SST.) Therefore we next focus on how the latent heat flux anomalies are generated.
Figure 9 shows the surface wind and latent heat flux
anomalies for each winter month in both the uncoupled
and the coupled experiments. We first focus on the
coupled results (Fig. 9, right). During December (Fig.
9b) the largest surface latent heating anomalies occur
west of the Sahara in the northern tropical Atlantic; at
and south of the equator the latent heating anomalies
are positive and weaker. A linear analysis, following
Saravanan and Chang (2000), indicates that the latent
heat flux anomalies are mostly a consequence of wind
speed anomalies (latent heat flux anomalies due to
changes in air–sea temperature difference are large off
the Saharan coast, but increase SST, the opposite of the
VOLUME 18
wind speed effect). The enhanced trades in the northern tropical Atlantic are a consequence of the insolation-induced cold anomalies and high surface pressure
over the Sahara (not shown). This conclusion is consistent with our analysis in section 4b below, which suggests that, for understanding the seasonal march of the
ITCZ, insolation-driven changes in continental temperature are more important than changes in continental convection.
The December latent heat flux anomalies extend to
the equatorial region where they force a negative meridional gradient in SST. In January (Fig. 9d), northerly
cross-equatorial wind anomalies are in place and cause
a strengthening of the meridional dipole in latent heat
flux anomalies in the equatorial region. By February
(Fig. 9f) the latent heat anomalies off northern Africa
have changed sign, while the negative cross-equatorial
gradient of SST is maintained, and intensified, by a
narrow patch of cross-equatorial wind. A comparison
with the uncoupled case (Fig. 9, left) shows that the act
of coupling amplifies the cross-equatorial wind and latent heat flux anomalies in the equatorial region (cf.
Figs. 9c and 9d and, especially, 9e and 9f). There is a
positive feedback between the wind and the SST: the
wind-induced latent heat anomalies strengthen the SST
gradient, which in turn drives a stronger crossequatorial wind. This wind–evaporation–SST feedback
was first introduced in an idealized study of the annual
cycle of the ITCZ by Xie and Philander (1994) and later
invoked to explain the tropical Atlantic decadal variability (Chang et al. 1997). Nevertheless, we note that
the latent heat flux anomalies in the uncoupled case
(Figs. 9a,c,e) are by themselves sufficient to induce a
negative SST gradient at the equator; in this sense, the
wind–evaporation–SST feedback, while undoubtedly
an integral part of the development and the maintenance in time of the SST gradient and the displacement
FIG. 8. The wintertime (the sum of Dec, Jan, and Feb) (left) total and (right) latent surface heat flux response
to the annual cycle of insolation over land in the case of interactive SST (CpldCTL ⫺ CpldLPVE). The contour
interval is 40 W m⫺2, starting with ⫾20. The thick line is the zero line of the Mar CpldCTL ⫺ Mar CpldLPVE
precipitation anomalies.
1 JANUARY 2005
BIASUTTI ET AL.
221
FIG. 9. The latent heat flux and surface wind response to the annual cycle of insolation over land in the case of
(left) prescribed and (right) interactive SST. The contour interval is 20 W m⫺2, starting with ⫾10; dashed contours
indicate negative values; and the shading indicates positive values larger than 10 W m⫺2. Only wind anomalies
larger than 1.5 m s⫺1 are plotted. (a) Dec CTL ⫺ Dec LPVE, (b) Dec CpldCTL ⫺ Dec CpldLPVE, (c) Jan
CTL ⫺ Jan LPVE, (d) Jan CpldCTL ⫺ Jan CpldLPVE, (e) Feb CTL ⫺ Feb LPVE, (f) Feb CpldCTL ⫺ Feb
CpldLPVE.
of the ITCZ in our coupled experiment, is not crucial
for either.
A complementary view of the development of the
March displacement of the ITCZ is portrayed in Fig. 10.
The top panel shows the annual cycle of the
CpldCTL⫺CpldPVE anomalies in SAT over the Sahara; these anomalies are directly forced by insolation
changes over land. The vertical gray line in Fig. 10a
highlights the coldest SAT anomaly during December.
(In the other panels of Fig. 10 the left side of the gray
bar is December, while the right side extends to the
time when the plotted anomalies reach their maximum
amplitude; thus the width of the gray bar gives a visual
estimate of the response time of each variable.) The
222
JOURNAL OF CLIMATE
VOLUME 18
cold anomaly in the Sahara produces a high sea level
pressure anomaly and easterly wind in the northern
tropical Atlantic, strengthening the trades (Fig. 10b)
and cooling the SST by evaporation. The maximum
cooling in the northern tropical Atlantic is achieved in
February (Fig. 10c) at which time virtually all of the
cross-equatorial SST gradient is also established (Fig.
10d). The ITCZ responds to the SST gradient by shifting south, achieving the largest southward displacement
in March (according to the confluence line, Fig. 10e) or
April (according to precipitation maximum, Fig. 10f).
From the above discussion we infer that the timing of
the maximum displacement of the ITCZ should be sensitive to the response time of the mixed layer, which in
this simple model is just proportional to the mixed layer
depth. We have performed a similar set of experiments
in which the depth of the mixed layer is reduced to 1 m
to make the ocean respond virtually instantaneously to
the atmospheric forcing; in that case the maximum
ITCZ displacement occurs in December (not shown).
We conclude this section with a caveat. In the above
discussion we have disregarded the state of the midlatitudes. Obviously the insolation forcing has a large impact on the temperature of the northern midlatitude
continents, and therefore on the midlatitude jets (but
not on the midlatitude ocean temperature because
those are prescribed in our model configuration). The
fact that the changes in tropical SST and precipitation
can plausibly be explained in terms of the insolationinduced temperature changes in the Sahara makes us
confident that the midlatitude jets play a secondary role
at best. Sensitivity experiments are under way in order
to verify that this is indeed the case.
b. The relative role of land precipitation and
land surface temperature in inducing an
ITCZ response
Figures 11a and 11d show the difference CpldCTL ⫺
CpldLPVE in the central Atlantic. As we have seen, the
anomalies in SST meridional gradient and surface wind
and the displacement of the ITCZ in boreal spring
←
FIG. 10. The response to the annual cycle of insolation over land
in the case of interactive SST (CpldCTL ⫺ CpldLPVE. (a) Surface air temperature in the Sahara. (b) Wind speed in the north
tropical Atlantic. (c) SST in the north tropical Atlantic. (d) SST in
the north tropical Atlantic and SST in the south tropical Atlantic.
(e) Position of the Atlantic ITCZ as measured by the confluence
line. (f) Position of the Atlantic ITCZ as measured by the maximum precipitation. The gray vertical line indicates the elapsed
time between the time of the minimum surface temperature in the
Sahara, Dec, and the time when the other indices reach an extrema. The wind response in the northern tropical Atlantic is
instantaneous. The minimum SST in the tropical Atlantic is
reached within 2 months, in Feb. The maxima in cross-equatorial
SST gradient and in the displacement of the ITCZ is reached in
Mar–Apr.
1 JANUARY 2005
BIASUTTI ET AL.
223
FIG. 11. The annual cycle of the response of precipitation, SST, and surface wind in the central Atlantic (30°W) to annually varying
continental forcing as a function of month and latitude. (a), (d) CpldCTL ⫺ CpldLPVE: anomalies due to insolation over land. (b), (e)
CpldCTL ⫺ CpldPMQ: anomalies due to precipitation over land. (c), (f) CpldPMQ ⫺ CpldLPVE: anomalies due to surface temperature over land. The contour interval for precipitation anomalies is 6 mm day⫺1, starting with ⫾3; dashed contours indicate negative
values; and shading indicates positive values larger than 3 mm day⫺1. The contour interval for SST anomalies is 0.5°C, the zero line is
omitted, dashed contours indicate negative values, and shading indicates positive values. Only surface wind anomalies larger than 2 m
s⫺1 are plotted; the bold arrows in the lower-right corner represent wind speed of 4 m s⫺1.
[March–May (MAM)] characterize the remote response over the Atlantic to seasonal changes in TOA
insolation over land. Both land surface temperature
and land precipitation changes due to the seasonal cycle
in land insolation are concurrent with the oceanic
anomalies portrayed here.
We can attempt to further decompose the effect of
land on oceanic precipitation into the effect of elevated
heating from land precipitation and the effect of land
surface temperature by comparing the CpldLPVE and
CpldPMQ experiments with the coupled control
CpldCTL. In the CpldPMQ simulation, land insolation
and ocean forcings (insolation and Q flux) cycle
through their climatology, whereas the elevated condensation heating associated with precipitation over
Africa and South America is kept fixed at March value.
Therefore, the difference plots CpldCTL ⫺ CpldPMQ
(Figs. 11b,e) show the impact of elevated heating
changes over Africa and South America (due to the
annual cycle of insolation over land) on the climate of
the central Atlantic. To the extent that changes in the
midlatitudes climate can be neglected and that the effect of land surface temperature and land convective
heating add up linearly, the difference CpldPMQ ⫺
CpldLPVE (Figs. 11c,f) can be thought of as showing
the maritime response to land surface temperature in
Africa and South America.2
Figure 11 shows that the CpldCTL ⫺ CpldPVE precipitation and SST differences (Figs. 11a,d) are the result of large but opposite responses to changes in the
elevated heating over the continents and to changes in
the surface temperature of the continents, with surface
temperature being the dominant forcing (cf. Figs. 11b
with 11c and Figs. 11e with 11f). Land elevated heating
induces a basinwide SST dipole anomaly, with a positive anomalous meridional gradient in SST and a northward shift of the ITCZ. Land surface temperature induces a negative anomalous meridional gradient of SST
across the latitude of the ITCZ to which corresponds
the southward movement of the ITCZ; outside of the
2
Note that the land surface temperature changes do include the
changes due to precipitation (and thus cloud cover and soil moisture): the land surface temperature changes are not those that
insolation changes would produce in a dry model. We cannot
separate out the total effect of continental precipitation changes.
What we are attempting to do is to distinguish how much of the
remote response to land climate is exerted though the free troposphere (i.e., in response to continental elevated heating) and
how much through the boundary layer (i.e., in response to surface
temperature).
224
JOURNAL OF CLIMATE
equatorial (ITCZ) region, land surface temperature
produces negative anomalies. We remark that the
ITCZ is consistently shifted north (south) throughout
the year in response to forcing from land elevated heating (surface temperature) that changes sign over the
course of the year.
There are two reasons for this. The principal reason
is that the forcings have a nonzero annual mean. For
example, the CpldLPVE is warmer in the mean than
the CpldCTL, which means that the forcing from the
CpldCTL ⫺ CpldLPVE land temperature during, say,
June does not compensate for that during December
(see Fig. 10); thus the annual mean response of the
ITCZ is biased toward the December value. A secondary reason is that the feedbacks among the SST, the
ITCZ, and the surface winds tend to maintain an original displacement of the ITCZ. Note that nonzero annual mean anomalies in SST are produced during the
adjustment period (by definition, the annual mean heat
fluxes are zero at equilibrium). Thus we cannot diagnose the mechanisms that gave rise to the aforementioned pattern of SST anomalies. Nevertheless, we
speculate that the same mechanisms highlighted in sections 3 and 4a should be responsible for bringing the
influence of land elevated heating and surface temperature, respectively, to the Atlantic Ocean.
VOLUME 18
5. The ITCZ response to annually varying
local forcings
In this section we separate the role of the annual
cycle in insolation over the ocean and of the annual
cycle in the Q flux (representing the ocean heat transport) in generating the annual cycle of the tropical
Atlantic ITCZ. We show results from two additional
experiments (see Table 2 and the appendix). In the
CpldLPVEQflux experiment, the only annually varying
forcing is insolation over the ocean; insolation over
land is fixed at the vernal equinox value and the Q flux
is prescribed to be the annual average of the Q flux
from the control integration. In the CpldLOPVE experiment, the Q flux has an annual cycle; insolation is
fixed at the vernal equinox value everywhere.
Figure 12 shows the annual cycle of precipitation,
SST, and surface winds in the CpldLPVEQflux and
CpldLOPVE simulations. Compared to the control
(Fig. 6a), the annual cycle in the position of the ITCZ
is reduced in both simulations (especially in the
CpldLOPVE in which the annual variations derive entirely from the Q flux). In both simulations, the equilibrium state has a year-round ITCZ in the north and a
second maximum of precipitation in the southern Tropics that lasts only few months. Thus, these experiments
would suggest that, while annual variations in both
FIG. 12. The annual cycle of precipitation, SST, and surface wind in the central Atlantic
(30°W) in simulations with different annual forcings. (a), (c) CpldLPVEQflux: annual cycle
due to local insolation. (b), (d) CpldLOPVE: annual cycle due only to the Q flux. The contour
interval for precipitation is 3 mm day⫺1, starting with ⫾1.5; values larger than 4.5 mm day⫺1
are shaded. The contour interval for SST is 1°C; values larger than 27°C are shaded. Only
surface wind vectors larger than 5 m s⫺1 are plotted; the bold arrow in the lower-left corners
indicates a 10 m s⫺1 wind speed.
1 JANUARY 2005
BIASUTTI ET AL.
ocean heat transport and insolation do generate a substantial annual cycle in the intensity of maritime precipitation, they have a weaker effect over the annual
meridional march of the ITCZ.
Yet, some caution in the interpretation of these results needs to be exercized. One caveat derives from the
fact that the annually averaged insolation is different in
the two simulations. In particular the northern Tropics
receive in the annual mean more insolation in perpetual
vernal equinox conditions then they do for climatological insolation. This additional insolation contributes to
the pronounced warming at about 7°N seen in Fig. 12d.
A second consideration we wish to emphasize is that
the above results are only valid in the context of
AGCM⫹SOM experiments. While it is appropriate to
refer to the Q-flux forcing as a local forcing in the context of AGCM⫹SOM experiments, in reality the ocean
heat transport that we parameterize with the Q flux is
the result of ocean dynamics that is, in principle, the
result of both local and remote forcings. Sorting out the
cause of annual variations in ocean heat transport requires a dynamical ocean, and cannot be done in our
simpler model configuration. Furthermore, the SOM
uses a constant, uniform mixed layer depth of 50 m. In
the real world, the depth of the mixed layer varies in
space and time, introducing another modulation in the
response of the ocean to surface fluxes. The depth of
the mixed layer is determined by the the strength of the
wind and by advective processes in the ocean, and thus
depends in a nontrivial way on both local and remote
forcings. Finally, although we have interpreted the Q
flux as a parameterization of the ocean heat transport
convergence, it also corrects for model biases, which
might present a nontrivial annual cycle.
6. Summary and discussion
We have investigated how changes in surface temperature and precipitation over South America and Africa affect the Atlantic ITCZ. We used the Community
Climate Model version 3 (CCM3) AGCM, coupled in
the tropical Atlantic to a slab ocean model (SOM) of
uniform and constant depth. Thus, the tropical Atlantic
SST can respond to radiative and turbulent fluxes but
not to the dynamical consequences of wind stress. The
effect of ocean dynamics is parameterized by a flux
correction (referred to as the Q flux).
In this model configuration the climate over the
tropical Atlantic Ocean is forced locally by the insolation overhead and the Q flux and remotely by the circulation driven by surface temperature and precipitation over land. We have performed experiments in
which these different forcings were applied independently to the CCM3⫹SOM model (see Table 2 for a
complete list of the coupled experiments presented in
this paper). Comparison of the annual cycle in these
experiments provides us with an estimate of how im-
225
portant each forcing is in determining the annual cycle
of the Atlantic ITCZ.
The main conclusion of this study is that variations in
precipitation and temperature over the continents are
as important as variations in insolation over the ocean
and in ocean heat transport convergence in forcing the
annual march of the Atlantic ITCZ observed in the
control simulation.
In sections 4 and 5 we have shown that the simulated
control climatology of the Atlantic ITCZ is the result of
a balance between the annual variations of local insolation, ocean heat transport, continental surface temperature, and continental precipitation, and that no
single forcing is dominant. The annual variations in insolation over the ocean and in ocean heat transport
modulate the intensity of precipitation in the tropical
Atlantic, both north and south of the equator, but neither is sufficient to force the cross-equatorial migration
of the ITCZ seen during boreal spring in the control
simulation. In contrast, the annual variations in terrestrial forcings (continental surface temperature and precipitation) affect the position of the ITCZ more than its
intensity, with the forcing due to continental precipitation (elevated condensation heating) partially counteracting the dominant effect of the continental surface
temperature forcing.
Figure 13 offers a succinct summary of our results. It
shows the amplitude and phase of the annual harmonic
of precipitation and surface temperature in response to
annually varying insolation and ocean heat transport
(CpldCTL; Figs. 13a,b), land insolation only
(CpldCTL⫺CpldLPVE; Figs. 13c,d), ocean forcings
only (CpldLPVE; Figs. 13e,f), ocean insolation only
(CpldLPVEQflux; Figs. 13g,h), and Q flux only
(CpldLOPVE; Figs. 13i,l). The length of each vector
gives a measure of the amplitude of the annual cycle
(more precisely, it represents the amplitude of the annual harmonic of the climatology) at that location, and
the direction represents its phase, with an arrow pointing upward indicating a maximum in January, and time
progressing clockwise.
A comparison of the annual harmonic of temperature and precipitation in the control run is interesting in
its own right: it shows that the largest amplitude in the
annual variations of oceanic precipitation in this model
is achieved in the western equatorial basin where the
annual harmonic of SST is negligible. SST has a much
larger annual harmonic in the eastern equatorial region
in the Gulf of Guinea. This local “decoupling” of the
SST and the ITCZ is even more apparent in the annual
harmonic simulated in response to the annual cycle of
insolation over land: The ITCZ response is very strong
(and strikingly similar to the control annual harmonic
in the equatorial region), while the SST response is
weak. The remotely forced annual harmonic of the
ITCZ is the result of a cancellation between the response to continental precipitation (Fig. 13m) and continental surface temperature (Fig. 13o). The locally
226
JOURNAL OF CLIMATE
VOLUME 18
1 JANUARY 2005
227
BIASUTTI ET AL.
forced annual harmonic in oceanic precipitation (i.e.,
that forced by insolation over the ocean and by the Q
flux) is of the same amplitude of that forced remotely
(cf. Figs. 13c and 13e) and is the result of partial cancellation between the effect of the insolation and the Q
flux (cf. Figs. 13g and 13i).
The annual harmonic of SST appears to be mainly a
response to local insolation, although variations in Q
flux are also important, especially in the Gulf of
Guinea. At first glance, this result seems at odds with
Carton and Zhou (1997), who conclude that insolation
had little effect between 5°S and 10°N. That they came
to a different conclusion may simply be due to the differences in the experimental design. Carton and Zhou
(1997) use an uncoupled ocean model, which they drive
with either the annual cycle or the annual mean of the
observed surface heat fluxes and winds. Thus, their experiment can only capture the uncoupled response of
SST to the annual cycle of insolation. However, in the
equatorial Atlantic wind, precipitation, and SST are
governed by strongly coupled dynamics.
Our conclusion about the role of ocean heat transport in determining the annual cycle of SST agrees with
that of Carton and Zhou (1997) qualitatively, but not
quantitatively. This might be a consequence of the fact
that in our Q-flux-only experiment there are significant
changes to the surface wind—and thus surface heat
flux—due to atmosphere–ocean coupling. Moreover,
extrapolating our conclusion on the relative role of annual variations in the Q flux in generating the annual
cycle of SST to the role of ocean dynamics is probably
not warranted: the Q flux also corrects for the choice of
a uniform constant mixed layer depth, and is contaminated by biases in the atmospheric fluxes.
In regard to the Q flux we also note that, in this
model configuration, the Q flux can be interpreted as a
local forcing, but in reality the ocean heat transport
convergence contains the effect of horizontal advection
by ocean currents, upwelling, and the deepening and
shoaling of the mixed layer depth. Thus, it is the product of local and remote forcings alike. In particular,
Mitchell and Wallace (1992) in an observational study
and Li and Philander (1997) in a modeling study have
suggested that the southerly wind in the (land driven)
African monsoon is the main driver for the dynamics of
the Atlantic cold tongue. A recognition of the total
effect of the continents on the oceanic climate requires
the use of a dynamical ocean model and is beyond the
scope of this study.
We conclude with a brief remark on the implications
of this study. We have shown that variations in the
African climate easily trigger a coupled response in the
surface wind, the ITCZ, and the cross-equatorial SST
gradient in the tropical Atlantic, whose net effect is the
meridional displacement of the ITCZ. Thus, good simulation of tropical Atlantic climate in interactive ocean–
atmosphere models will need good representation of
processes over the neighboring continents and their
connections to the oceanic sector.
As an interesting aside, we not that our results on the
influence of the Sahara on the Atlantic suggest that
paleoclimate changes like the mid-Holocene greening
of the Sahara (with the accompanying changes in surface temperature and precipitation) should be traceable
in paleo records of the ITCZ position in the equatorial
Atlantic. Furthermore, our results suggest that—insofar
as the mechanisms responsible for interannual variability are indeed similar to those responsible for the annual cycle as it has been claimed (Hastenrath 1984)—
variability in, say, African climate might contribute to
the variability of the ITCZ.
Acknowledgments. The authors wish to thank an
anonymous reviewer, Mike Wallace, Dennis Hartmann, and especially John Chiang for their insightful
suggestions and editorial comments. This publication is
supported by a grant to the Joint Institute for the Study
of the Atmosphere and Ocean (JISAO) under NOAA
Cooperative Agreement NA17RJ1232.
APPENDIX
Experiment Summary
a. Uncoupled simulations
• CTL: The control simulation is run with prescribed
annual cycle of SST and top-of-atmosphere (TOA)
insolation.
• PMS: The perpetual March SST simulation has SST
fixed at March value, but climatological TOA insolation. The monthly condensational heating over
South America and Africa (hereafter referred to as
Q) is saved and used to force other simulations.
• PM: The perpetual March simulation has March SST
and 21 March equinox insolation. Thus PM will be
←
FIG. 13. Harmonic dial of the annual cycle of (left column) precipitation and (right column) surface air temperature
in (a) and (b) CpldCTL; (c) and (d) CpldCTL ⫺ CpldLPVE; (e) and (f) CpldLPVE; (g) and (h) CpldLPVEQflux;
(i) and (l) CpldLOPVE; (m) and (n) CpldCTL⫺CpldPMQ; (o) and (p) CpldPMQ⫺CpldLPVE. The direction of the
arrows indicates the max in the annual harmonic, with Jan pointing upward and time increasing clockwise. The length
of the arrows indicates the amplitude of the annual harmonic. The smallest arrow in the precipitation plots indicates
an amplitude of 0.5 mm day⫺1. The smallest arrow in the surface air temperature plots indicates an amplitude of
0.25°C.
228
JOURNAL OF CLIMATE
similar, but not identical, to March conditions in CTL
and PMS.
• PMJQ: This simulation is a perpetual March, like
PM, but Q is prescribed to its June value. In short,
PMJQ is PM with heating added as in Fig. 1.
• LPVE: The land perpetual vernal equinox simulation
has 21 March TOA insolation over land and climatological TOA insolation over the ocean. Because the
SST is prescribed, LPVE is virtually identical to PVE
in BBS1 in which insolation is fixed over both the
land and the ocean. Because the precipitation over
land is largely determined by insolation, the condensational heating over land is very similar in LPVE,
PM, and March conditions in CTL.
b. Coupled simulations
• CpldCTL: The control simulation is run with prescribed annual cycle of TOA insolation and annual
cycle of Q flux. (The Q flux, that is, the flux correction required by the slab ocean model to reproduce
the observed SST climatology, is determined using
the surface heat fluxes from CTL.)
• CpldPMQ: The perpetual March heating is like
CpldCTL, but Q is prescribed to its March value
(from PMS).
• CpldPJQ: The perpetual June Q is like CpldCTL but
with prescribed June Q. In short, CpldPJQ is
CpldPMQ with added heating as in Fig. 1.
• CpldLPVE: The coupled land perpetual vernal equinox simulation has 21 March TOA insolation over
land and climatological insolation and Q-flux forcing
the ocean. To the extent that precipitation over land
is largely determined by insolation, the condensational heating over land corresponds to March conditions.
• CpldLPVEQflux: This has 21 March TOA insolation
over land, climatological insolation over the ocean,
and annual-mean Q flux. To the extent that precipitation over land is largely determined by insolation,
the condensational heating over land corresponds to
March conditions.
• CpldLOPVE: This is forced by 21 March TOA insolation everywhere, over both land and ocean, and by
the annual cycle of the Q flux. To the extent that
precipitation over land is largely determined by insolation, the condensational heating over land corresponds to March conditions.
REFERENCES
Biasutti, M., D. S. Battisti, and E. S. Sarachik, 2003: The annual
cycle over the tropical Atlantic, South America, and Africa.
J. Climate, 16, 2491–2508.
VOLUME 18
——, ——, and ——, 2004: Mechanisms controlling the annual
cycle of precipitation in the tropical Atlantic sector in an
atmospheric GCM. J. Climate, 17, 4708–4723.
Carton, J. A., and Z. Zhou, 1997: Annual cycle of sea surface
temperature in the tropical Atlantic Ocean. J. Geophys. Res.,
102 (C13), 27 813–27 824.
Chang, P., L. Ji, and H. Li, 1997: A decadal climate variation in
the tropical Atlantic Ocean from thermodynamic air–sea interactions. Nature, 385, 516–518.
——, R. Saravanan, L. Ji, and G. C. Hegerl, 2000: The effect of
local sea surface temperatures on atmospheric circulation
over the tropical Atlantic sector. J. Climate, 13, 2195–2216.
Chiang, J. C. H., S. E. Zebiak, and M. A. Cane, 2001: Relative
roles of elevated heating and surface temperature gradients
in driving anomalous surface winds over tropical oceans. J.
Atmos. Sci., 58, 1371–1394.
——, M. Biasutti, and D. S. Battisti, 2003: Sensitivity of the Atlantic ITCZ to Last Glacial Maximum boundary conditions.
Paleoceanography, 18, 1094, doi:10.1029/2003PA000916.
Davey, M. K., and Coauthors, 2002: STOIC: A study of coupled
model climatology and variability in tropical ocean regions.
Climate Dyn., 18, 403–420.
Gill, A. E., 1980: Some simple solutions for heat-induced tropical
circulation. Quart. J. Roy. Meteor. Soc., 106, 447–462.
Hack, J., J. Kiehl, and J. Hurrell, 1998: The hydrologic and thermodynamic characteristics of the NCAR CCM3. J. Climate,
11, 1207–1236.
Hastenrath, S., 1984: Interannual variability and the annual cycle:
Mechanisms of circulation and climate in the tropical Atlantic sector. Mon. Wea. Rev., 112, 1097–1107.
——, and L. Greischar, 1993: Further work on the prediction of
northeast Brazil rainfall anomalies. J. Climate, 6, 743–758.
Kiehl, J. T., J. J. Hack, G. B. Bonan, B. A. Boville, D. L. Williamson, and P. J. Rasch, 1998: The National Center for Atmospheric Research Community Climate Model: CCM3. J.
Climate, 11, 1131–1150.
Li, T., and S. G. H. Philander, 1997: On the seasonal cycle of the
equatorial Atlantic. J. Climate, 10, 813–817.
Lindzen, R. S., and S. Nigam, 1987: On the role of sea surface
temperature gradients in forcing low-level winds and convergence in the Tropics. J. Atmos. Sci., 44, 2418–2436.
McGauley, M., C. Zhang, and N. A. Bond, 2004: Large-scale characteristics of the atmospheric boundary layer in the eastern
Pacific cold tongue–ITCZ region. J. Climate, 17, 3907–3920.
Mitchell, T. P., and J. M. Wallace, 1992: The annual cycle in
equatorial convection and sea surface temperature. J. Climate, 5, 1140–1156.
Saravanan, R., and P. Chang, 1999: Oceanic mixed layer feedback
and tropical Atlantic variability. Geophys. Res. Lett., 26,
3629–3632.
——, and ——, 2000: Interaction between tropical Atlantic variability and El Niño–Southern Oscillation. J. Climate, 13,
2177–2194.
Seager, R., Y. Kushnir, P. Chang, N. Naik, J. Miller, and W.
Hazeleger, 2001: Looking for the role of the ocean in the
tropical Atlantic decadal climate variability. J. Climate, 14,
638–655.
Xie, S. P., and S. G. H. Philander, 1994: A coupled ocean–
atmosphere model of relevance to the ITCZ in the eastern
Pacific. Tellus, 46A, 340–350.